How do companies create value with AI?

Companies create value with AI by changing how work gets done, not by layering tools onto the work they already do. The payoff is often faster cycles and lower costs. But it shows up only when leaders treat AI as a reason to redesign workflows, roles, and decisions, not to automate the status quo.

Here’s why that matters. According to a Bain survey, nearly 40% of companies that measured AI cost savings landed below 10%, despite targeting 11% to 20%. AI investments stall not because the technology fails, but because it gets dropped into processes before anyone stops to rethink them.

The companies on the right side of this performance gap didn’t find the best AI technology. They made specific organizational decisions. Those capturing real value redesign the work before they apply the tool. They don’t ask, “Where can we apply AI?” They ask, “If we were designing this process from scratch today, what would it look like?” Only then do they turn to the technology.

Why aren’t most companies capturing AI's value yet?

Most companies aren’t capturing AI’s value yet because they settle for micro-productivity—faster reports, fewer coders, modest task-level gains. Too many treat it as a way to do the same work with less effort, not an engine to do more. If technology’s impact on people is handled as a downstream change-management problem, AI won’t move past those small wins. The result is disappointing ROI, disengaged employees, and growing skepticism.

Spending is up. Returns aren’t. Despite falling short on measured AI cost savings, 90% of companies are raising their AI budgets again, according to our survey. Few are pausing to understand why AI is underdelivering on their targets.

The pattern repeats across waves—robotic process automation, then machine learning, then generative AI. Now, it’s agents that will operate with even greater autonomy, complexity, and consequence. The shortfall is rarely dramatic enough to kill a program. But it’s quiet, consistent, and large enough to matter to executives.

The companies breaking the pattern didn’t find better technology or bigger budgets. They treated AI as a test of leadership and vision, not just another technology cycle. They simplified and standardized their workflows while modernizing their workforce in parallel, not treating people as a downstream fix.

Where does AI’s value come from?

AI’s value comes from three reinforcing sources: speed, cost, and productivity. At scale, the three together reset how fast a company can operate and how much it can do with the resources it has.

However, companies that bolt AI onto existing work capture only a sliver of the value. It’s the companies that redesign work around AI that capture the structural gains.

Speed

Speed is the first dividend AI pays: When finance leaders name their biggest AI win, speed and cycle-time reduction lead at 48%, ahead of headcount or cost savings at 34%, according to a Bain CFO survey. Tighter close cycles, faster reconciliations, and earlier variance insight let a company detect problems, correct course, and redeploy capital sooner.

The strategic value runs deeper than the metrics suggest. When trade policy shifts, rates swing, or supply chains break, the ability to reforecast quickly, reallocate capital on short notice, and surface risk in real time is a competitive advantage, not just an operational one. Compressing the path from market signal to decision from weeks to days helps the business move faster than its rivals.

Leading CFOs will treat speed as a strategic outcome and measure it directly—days-to-close, forecast cadence, and time-to-variance resolution—with the same rigor as expense. When speed is the headline metric, AI’s real value—better decisions and faster course correction—becomes visible and defensible.

Cost

AI is a powerful tool for taking out costs, including the cost of AI itself. Built to perform complex, knowledge-intensive work faster and more efficiently, it can surface hidden spending, identify underused software, streamline operations, and optimize infrastructure.

AI brings near-real-time visibility into IT costs, flagging shadow spending that sits beyond a tech leader's line of sight. It rationalizes bloated application portfolios—pruning that typically cuts software and maintenance costs by 10% to 30%. Generative AI coding assistants and automated testing tools shrink software development cycles by 20% to 30%, with lower labor costs, better quality, and faster time-to-market.

But there’s a catch. AI also adds cost and complexity. Nearly 70% of tech leaders expect it to raise spending by more than 5%. Beyond direct costs, AI increases the complexity of running a business, through faster tech cycles, retooled architecture, and the development of new operating models. When layered onto already-fragmented ecosystems and aging core systems, those models and agents create new integration challenges and, in some cases, higher operating costs.

Leading IT and transformation teams build a flywheel. They fund AI with AI, using its efficiencies to offset the cost of broader adoption. One powerful example is a tiered model strategy: smaller, fine-tuned models handle high-volume routine tasks at a fraction of the cost, while reserving large models for complex, high-impact work.

Productivity

AI has exponential productivity potential, but only when companies modernize workflow and workforce in parallel. Workflow modernization means deep process reengineering and simplification. Workforce modernization means smarter teaming, strategic workforce planning, and continuous reskilling. The two are inextricably linked, yet too often, workforce change trails behind workflow redesign. That lag is where value leaks away.

Firms that link the two are seeing a 10% to 15% productivity lift depending on the domain, translating to 10% to 25% EBITDA gains that grow as programs scale.

Five steps can synchronize workflow and workforce modernization to boost productivity: prepare leaders and teams, avoid automating workflow debt, shift to a hybrid workforce, build trust, and capture value with new performance standards.

• Prepare leaders and teams. Cross-functional teaming of technology, business process owners, finance, and HR is essential.
• Don't automate workflow debt. Most companies carry workflow debt—the accumulation of unnecessary meetings, approvals, handoffs, and one-off exceptions that pile up around even simple tasks. AI amplifies whatever system it enters. Humans can work around fuzzy rules; agents can't. Deploy AI on top of workflow debt, and it multiplies complexity instead of productivity. Getting process and work design right is a prerequisite to scaling agents, not a cleanup job for later. Organizations that get real returns begin with a clean sheet for their most critical processes, setting bold targets for customer experience, cost, and speed. Then they work backward to designs that deliver those targets. They strip out low-value work and decide what humans must do, what humans should do with AI’s help, and what can run autonomously.
• Prepare a hybrid workforce. The future workforce blends four kinds of workers: people using off-the-shelf AI tools, “super humans” made far more productive by custom tools, autonomous and semi-autonomous agents, and humanoid robots. Humans will focus where they add the most value—relationships, mentorship, innovation, and complex judgment—while agents and robots take on more “run the business” work.
• Build trust. Employees need to believe the company is investing to augment—not replace—them. That means reskilling, redeployment, and clear communication of intent.
• Capture the value and set new standards of performance. Embed new workflows in enterprise management systems and technology platforms, as well as new labor staffing rules for strategic workforce planning.

Done well, this becomes a perpetual productivity engine, led by humans and powered by AI, where people and machines keep learning from each other. Companies that pair high workforce engagement with high productivity deliver total shareholder returns 2.3 times those that don't.

Why scaling AI - not just using it - separates the value leaders

The companies that get the most value from AI won’t be the ones that simply use it; they'll be the ones that scale it. So far, few have. Quarter after quarter, Bain’s survey of enterprises tells the same story: Fewer than 20% have scaled their generative AI efforts in any meaningful way.

Companies that move AI out of pilots and into production report greater satisfaction with the results. For instance, among CFOs still piloting, just 25% call their AI outcomes strongly positive. But among those who have scaled it into full production, 41% do. For the most AI-mature organizations, it exceeds 60%. The return on AI hinges less on how much you spend than on how far you scale.

The emerging leaders focus on a few high-value domains and redesign processes with AI at the core to drive scale and ROI. That means doing the hard work upfront: defining the right domains, setting top-down value hypotheses, and building the mechanisms to measure, manage, and scale the change over time.

What stands in the way of creating value with AI?

Three obstacles can stand in the way of creating value with AI: workflow debt, data access, and the autonomy gap.

Workflow debt

The most expensive mistake in AI is automating a broken process. AI doesn’t fix workflow debt. It locks it in, speeds it up, and makes it costlier to unwind. Before any AI program, leaders will consider how they can redesign processes from scratch.

Data

Data access and integration is the single biggest barrier to AI progress, cited by 41% of companies—ahead of compliance, budget, skills, and executive buy-in—in a Bain survey of 951 global companies. The strongest performers cite data as a bigger obstacle than those that missed their targets, because they’re deploying at scale and hitting the wall harder. The data problem is real. Using it as a reason to wait is not.

The autonomy gap

Most investment cases assume full automation economics. But only 7% of companies run fully autonomous agents in production today. The dominant model—cited by 38% of companies—still requires human approval. Another 32% run with guardrails and exceptions. Human oversight is the right posture. But when a business case is sized on full autonomy and the reality routes work to a human queue, CFOs have approved one set of numbers while the organization lives with another.

How can leaders create durable value with AI?

Leaders who turn AI investment into value make a handful of specific organizational decisions.

• Pay down workflow debt before deploying AI. Map the handoffs, approvals, and exception paths in high-priority processes. Simplify decision rights and stabilize rules before introducing autonomous execution. Do this first, and AI agents have the clear rules, stable handoffs, and decision rights they need to scale and earn trust. Skip that step, and they push work back to people or erode it.
• Validate the business case and name an accountability owner. Audit the actual returns of prior automation, not what was projected. A Bain survey shows 44% of companies are funding the next AI wave from automation savings that came in below target. Leading CEOs will answer the one question IT can’t answer for them: “Who’s personally accountable when an AI agent makes a consequential wrong decision?” Establishing that costs nothing and takes an afternoon.
• Use AI to attack the data problem; don't wait for it to be solved. Start where the data is already bounded and accessible. Save the large-scale data modernization conversation for the use cases that genuinely require it, and let early wins pay for it.
• Redesign the operating model, not just the process. Deploying AI agents without changing how people work around them is a fast track to under-delivering. In an agent-led model, people orchestrate, supervise, and make the calls agents can't. It’s a genuinely different role that requires deliberate investment in role redesign and change management.
• Measure value at the enterprise level. Programs optimize for what they measure, usually cost and hours saved. But it also matters whether AI is producing better decisions, faster responses, and stronger customer outcomes. If those metrics aren’t on the CEO’s dashboard, programs will keep delivering the wrong things efficiently.

The turning point is rarely a better model or a bigger budget. It’s the moment leaders decide they’re personally responsible for creating the conditions in which AI can succeed and generate value.